Marconi TF144H CT452A RF Generator

[Radio Wrangler]

I think the reason for the peak-reading detector for the ALC loop was that they were applying AM via the ALC loop. This is common in many lab sig gens in order to use the loop to linearise the modulation. Having the separate mean reading meter makes the whole thing a bit of a dog's breakfast.

Later sig gens really do control the carrier power via fast ALC and apply linearised amplitude modulation using the one loop and one detector and they get it sufficiently accurate that there is no need to hang a meter on it.

David

[vintage_8bit]

I think its probably a speed issue with feedback coming from the thermocouple. But it's impressively responsive when adjusting the output level. I wonder what the originals were like without the vacuum.
Can anyone explain how the A.L.C. monitor works. Why 2 diodes 2 caps and the feed from the HT via R193, 192,& 101. (A radio can get away with just one diode for the AGC.) I wondered if MR101 was there to provide some DC feedback to reduce overvoltage conditions during warm up?

[Radio Wrangler]

without the circuit in front of me, I can't remember, but it may be a system to provide a little compensation for diode forward voltage, and so improve the linearity at the bottom end.

David

[trh01uk]

Folks,

I too would like to hear some explanation of how the thing works, so here is a copy of the schematic for the RF system, which includes the metering, the ALC loop and the RF oscillator.

One other odd thing is that the ALC loop doesn't seem to actually stabilise the output at a set level. If you change frequency much, then you have to reset the level manually. Perhaps this is a problem with the flatness of the ALC detector?

Richard

[trh01uk]

And I might add that the manual is rather vague on the details of this circuit. For instance, what is the neon V104 actually doing? I can't find a datasheet for a type "3L", however I thought neons needed about 90V across them to actually carry any current?

And yet the voltages marked on the schematic show 44V at the cathode of V102A and 25V at the screen grid of V101 - with the neon and various resistors in between. There is apparently DC current flowing, given the voltage difference, but how exactly is that arranged with a neon in circuit? Or this some peculiar type of neon with an extremely low sustaining voltage?


Richard

[Radio Wrangler]

OK,

RF

The RF side is quite simple

V101 is a power oscillator using a 5763 I remember in mine, so QV03-12 is an equivalent european designation I assume. It runs at a significant power level for a signal generator which allows it to drive several things via attenuators with enough attenuation to protect against pulling by load changes. Output goes via S100c, R104 to a coarse step attenuator then a fine step attenuator and the output. C124 gives a little boost at the top end of the frequency range to compensate for some roll-off in output.

The level of the oscillator itself is varied for fine output adjustment and for AM application, by variation of g2 voltage for the 5763. This isn't very nice. Modulating the oscillator itself will introduce residual FM, and the amount of negative excursion of the modulation must be limited or else the oscillator will stop. Later and better sig gens adopted a master oscillator - power amp architecture to separate frequency control and modulation functions.

The RF output also drives the thermocouple and thence the RF level meter. MR103 and MR104 are an attempt to protect the thermocouple (which history has shown isn't good enough protection) They form a clamp biased by the DC heater voltage. If there is too much RF, the diodes conduct, forming a voltage doubler rectifier which tries to drive the heater rail. The HP2800 is a fast 70v Schottky diode and the current rating/series resistance isn't up to the power that oscillator valve can run at. 5763s were used in some topband transmitter designs to run the full 10W licence power.

The thermocouple is used for the RF level meter only. It's job is to be slow and show the mean power only. Things have changed and people want to know the carrier power these days, because the mean power increases with increased modulation.

The RF output also drives a fast-responding detector purely for ALC/modulation control. MR101/MR102. More of the HP Schottkys.

R195 supplies a little current from the HT rail to forward bias the detectors a little to improve linearity at the low power end of their range.

ALC

R114, RV101, RV102, R115 set the initial DC conditions for the DC amplifier V102B.

THe ALC detector voltage sets g1 voltage of V102B

V102 amplifies and inverts the ALC detector voltage. More detector volts makes its anode voltage fall. V102A is just a cathode follower buffering the anode voltage so R113 can be large for high gain, and still give enough current to drive the oscillator g2 etc. R112 gives a bit of positive feedback to increase the gain further, AC feedback via C130 acts to keep the DC amplifier stable and to shape the response to give enough bandwidth to handle the modulation without losing stability. C130 makes the loop amplifier act as an integrator, R117 creates a zero in the response (a phase lead) and then at higher frequencies C151 creates a pole (phase lag)

The anode potential of V1032 isn't reduced enough by the cathode follower, so the neon bulb V104 is used as an additional fixed voltage dropper to put the output DC in the right vicinity for the 5763's g2.

Modulation

Modulation is AC coupled into the output of the ALC detector via C129 and R171. As the modulation tries to swing this node positive, the loop responds by reducing the oscillator amplitude so the ALC detector output falls in order to compensate for the positive drive on the ALC detector output. A pair of gangs of the range switch allow modulation to be scaled slightly differently, or filtered differently on diferent bands.


Oh and contact 4 on the turret allows different bands to insert a load resistor to change the ALC bias if necessary.

Well, that's the walk-through done. Thanks for posting thr diagram.

David

Edit... written before your second posting which must have appeared while I was still typing. The ALC gain isn't enough, really, to give uniform operation between bands so you wind up twiddling RF level all the time, but it's as much gain as they could keep stable.

The HP 606A and 608A generators are much better behaved Master osc- PA designs with flat performance, fewer controls. The crazy thing is that they're simpler overall. Marconi just kept adding compensators to their compensators. The folk at Stanford got the basic system better and avoided all the kerfuffle.

I've covered the neon.

[G0HZU_JMR]

My knowledge of valves is limited to the triode and I've never used (let alone worked on) this TF144 sig gen, and I only have the schematic to go by but I can give you a few first impressions...

This looks to be some kind of old school way of providing ALC that is independent of the modulation. i.e. In this system the ALC needs to be blind to the modulation and therefore the loop bandwidth of the ALC system will be very low, perhaps just a few Hertz or less. So I don't fully agree with David's description because he implies the ALC system will have bandwidth wider than the modulation frequency. This would require an ALC bandwidth of several tens of kHz and the circuit doesn't support this in my opinion.

Someone asked about the design of the RF voltage detector in the ALC loop using the two diodes and it looks to me like this is an average reading detector. It (ideally) needs to be an average responding detector in order to help the ALC system reject the modulation. I think the initial design goal of the ALC detector will be that it will produce the same detector output for both a modulated and an unmodulated signal. It may not achieve this in reality but I think that is the idea at least as long as the modulation frequency is well up into the kHz region.

So this detector will be designed to (quite slowly) respond to the average voltage of the (un)modulated RF signal at the output of the sig gen. Looking at the detector circuit there is a div/100 from the HT and this will bias the detector diode MR102 to about 2.5V at a guess.

So I think the DC output of the detector measured at C160 will follow this crude equation:

Det out = ((Vpk-Vdrop) * 2/Pi) + 2.5 Volts

So for example, if the Vpk was 4V then the detector would produce ((4 -0.5) * 0.637) + 2.5 = 4.7V if the Vdrop in the MR102 diode is about 0.5V.

The 2/Pi (= 0.637) converts a 'complete' half sine to its average in theory. However, the 0.637 number needs to be a bit smaller to characterise the detector for the loss of part of the sine wave due the diode action. Therefore, the output will be slightly lower than the equation predicts. The second diode MR101 is likely to be there to provide a load for the negative half of the RF cycle.

I think the thermocouple is 'there' to provide a more accurate level detector and it will also be able to display/cal the modulation because it responds to the average power of the signal rather than the average voltage. So the thermocouple can crudely indicate modulation depth although you can probably only do this with reasonable accuracy at fairly high levels of modulation.

Do you precalibrate it for a high level of modulation on the meter and then turn it down from there? I've not used this generator so I can only guess.

The diode based ALC detector should be fairly immune to highish frequency modulation because it outputs a DC level via R107 R171 and C160 corresponding to 'average' RF voltage. So it won't be able to respond very well to modulation levels unless the mod frequency was quite low because of the long time constant of R107 and the total system capacitance around the node of C160.

Looking at the overall ALC system it will need the R and C components to be in good health in order to prevent instability. My guess is that instability will manifest itself at a very low frequency (a few Hz or less?) causing the RF level to wobble up and down. If there is greater than unity gain around the ALC loop at unity phase angle at the wobble frequency then the system will go unstable at the wobble frequency. You probably only need a few caps to age and lose value or go leaky or go high ESR for the loop gain to go up and you will lose your phase margin in the loop causing it to go unstable at this low wobble frequency. Even the resistors could cause problems if they age and the value changes.

You could draw the V102 valve as an error amplifier (i.e. crudely model it as an op amp?) to get some idea of the loop gain and phase response. The +ve input would be the reference (target) DC voltage for the ALC system which corresponds to pin 3 of the V102 valve and the -ve input would correspond to pin 2 of the valve.

The opamp output would correspond to pin 8 of the V102 valve. If you put all the RC components in you could at least get an idea of the loop BW of the system. I suspect it will be a few Hertz and this will mean that there will be a lower frequency limit for the modulation where the ALC will fight/cancel any attempt to modulate the generator. This is likely to be below 10Hz at a guess. But at modulation frequencies above this, the ALC and its detector will be blind to the modulation and the modulation will be successful. Try and think of how it is possible to modulate a PLL with FM (by injecting audio at the tune voltage location) as long as the modulation frequency is much higher than the loop bandwidth of the PLL. That's how a lot of FM radios work only in this case the loop is an AM based system rather than FM.

The best way to model the system would be to produce a working model for the V102 Valve used as the error amplifier rather than use an opamp but I don't know enough about valves to do this.

This info may not be completely accurate but then I've only looked at the circuit for a short time and I know very little about valves. Hope it helps

[G0HZU_JMR]

To fully characterise the whole system you would ideally need to break open the ALC loop and measure the 'loop gain' contribution of the oscillator in terms of how much the ALC detector output varies with respect to DC input to the level control pin of the oscillator valve.

This would be a fair bit of work and the whole system looks like it will not respond well to ageing because there are so many parts of this circuit that are variable. It looks like there are fudge factors switched into the loop according to range and this is presumably there to offset the variation in 'loop gain' caused by the varying behaviour of the oscillator across the various frequency bands. The oscillator would ideally produce the same RF level vs DC control voltage across all bands but in reality this won't be the case.

[Radio Wrangler]

I don't imagine there is enough loop bandwidth for broadband external AM, but if there isn't enough for basic 400Hz from the internal AF oscillator, then there are several inconsistancies.

ALC from an average detector would be OK, but a diode detector into a slow RC is going to be peak sensitive if it's slower than the modulation, so increasing % modulation will cause the ALC to depress the carrier power. To do an average detector would need a fast detector, an impedance buffer and then a slow averaging filter, else the troughs would not get averaged - the diodes would just be held off over mod troughs.

Screen-grid (well beam plates in the 5763) modulation is decidedly non-linear, so closed loop modulation would be very desirable in any lab generator used to test demod distortion in receivers.

The 2.5v bias, I now think isn't to grease-up the detector diodes, because all paths through the diodes lead to blocking capacitors or the ALC amp grid and besides there is the path through the two 1k resistors across the diodes. THe whole thing is floated at 2.5v to bias the ALC amp pentode.

This is really a last hoorah for old-school crude sig gens. It's a one-valver with a lot of sticking plasters over its cracks. Bizarrely all the additions take as much parts and money as doing it basically right in the first place. I imagine the designers being told to just tart the TF144G up a bit and put it in a modern looking box.

It's still well made and better than advance/taylor etc repair shop generators, but there was a new generation just around the corner.

David

[G0HZU_JMR]

I think that the ALC loop bandwidth is going to be a few Hz or less and I just looked up the specs for this sig gen and it allows up to 20kHz external modulation.

If you wanted to AM modulate via the ALC you would surely inject the AM at the reference pin of the error amplifier and this is the 'SET ALC' pin3 of the V102 valve. This would dynamically alter the ALC target and the ALC system (if it was fast enough) would effectively track this modulation target signal and it would effectively AM modulate the system in sympathy with the AF signal at the ALC reference pin.

That's how the later generation sig gens work.

But if you look at this generator the external or internal AM is injected into the other port (pin 2) of the error amplifier.

So the only way this can work is if the loop bandwidth is a few Hz or less and the typical modulation frequencies are much higher than this and the modulation becomes invisible to the ALC system. The problem with the ALC fighting the modulation will only occur down at very low mod frequencies that are close to the loop bandwidth of the system.

As far as the average voltage detector in the ALC loop is concerned the 2.5V dc bias level is just a pedestal voltage that it rides on (as shown in my equation). It could well be the required bias voltage for the valve but I have no idea what type of valve is used at V102.

[karesz*]

Jeremy,
V102 is originally type of 6U8 or may be replaced with CV5065/ECF82...
Regards, Karl

[G0HZU_JMR]

Radio Wrangler: ALC from an average detector would be OK, but a diode detector into a slow RC is going to be peak sensitive if it's slower than the modulation, so increasing % modulation will cause the ALC to depress the carrier power. To do an average detector would need a fast detector, an impedance buffer and then a slow averaging filter, else the troughs would not get averaged - the diodes would just be held off over mod troughs.

You are clearly seeing something different to me here David In my opinion the ALC detector is an 'average detector' in terms of it outputting a DC voltage at C160 that corresponds (via my crude equation given earlier) to the average voltage of the positive half of the RF signal. If LF sinewave modulation is applied at any modulation depth across 0 to 80% then the average voltage should remain virtually unchanged if we assume that there is symmetry in the modulated waveform.

Any change in the detector output across this range of modulation depths (assuming the modulated RF waveform has reasonably symmetric distortion if any appears) should be very small when compared to a 1dB change in overall RF level. i.e. the average reading detector won't be perfect but I think it will approximate to an average detector quite well indeed

That part of the system design has some elegance in my opinion because the ALC detector output should be largely unaffected by the modulation depth even at 80% modulation. However, I'm not sure what happens if you explore the lower frequency ranges of this generator. It looks like it works down to 10kHz. But at typical RF frequencies (LW/SW/VHF) and at modulation frequencies in the kHz region I think the detector will behave as an average voltage detector. So the ALC detector will not be able to report the modulation depth and this clashes with the suggestion that the generator works in closed loop for the modulation. This is because the average reading detector can't adequately detect that the modulation is there!

V102 is originally type of 6U8 or may be replaced with CV5065/ECF82...
Regards, Karl

Thanks Karl. Sadly my knowledge of valves ends with the basic triode so I think someone else will need to step in here. Those numbers mean little to me. I think the 6 has something to do with the heater voltage but that's about it

[G0HZU_JMR]

Maybe this will help wrt the way (I think) the ALC detector works.

If the design of the ALC detector was perfect (and we ignore the 2.5V pedestal offset in the TF144) it would try and give an output that corresponds to the average voltage of the RF cycle. See the waveform below.

Think of the blue waveform as a single RF cycle. The detector diodes and the RC integrator at the output will effectively produce an output that is 0.637 * Vpk in the case of an RF carrier that is a sinewave.

So this would be 1.91V DC However, in the real world there will be some Vdrop in the diode detector and this spoils things a bit. But it should still give something like (3-0.5V) * 0.637 = 1.59V.

In reality it will still be a bit lower because the 0.637 fudge factor applies to a complete half cycle of a sinewave but it should still be above 1.5V DC I think. Note that you would have to use a voltmeter with a very high input impedance to measure this DC voltage fairly accurately. Ideally much higher than 10Meg ohm. Don't use an AVO 8 here

Now we can consider what happens if the AM modulation is applied. The 'average' voltage contribution of a complete cycle of the low frequency modulating sine wave (to the RF carrier) is zero so the detector output won't change when the modulation is applied (in theory at least) This means the detector should be highly immune to modulation with a clean sine wave.

In the TF144 there is an additional 2.5V DC pedestal added to this detector output but that is less relevant for this analysis. I hope I've explained that well enough.

In the real generator a few things may spoil things a bit further. eg there may be high harmonic levels from the oscillator and also the AM modulation may be distorted enough to affect the shape/symmetry of the modulation envelope.

Hope all of this is OK. As I've said already, I'm not really a valve person so I don't fully undersand the V102 error amplifier and there may be some things here that I have overlooked.

[Radio Wrangler]

I've been looking at the 0.1uF capacitors and the 1k resistors and seeing 100us timeconstants in the detector... but the voltage pedestal input is soft with 100k source impedance.

No, I'll reserve further comment until I've chucked this one at Spice tomorrow!

David

[vintage_8bit]

To all.
Looking at the circuit in range "K" (One of the ranges I am having trouble with) I think the 2.5V potdown is shorted to ground through the coils of the adjacent "J" coil pack.

My fault.....
With the output set to 21mhz the "Dirrect Output" voltage is only 0.8V p to p. At 33mhz it rises to 10V p to p.
Looking at the output of the ALC at R107 / R171 I have around 1V DC. This doesn't change as the output frequancy and the p to p amplitude changes widely as above. This looks like an easy fault but have not found anything obvious yet.
Thanks for the previous replies I will try and absorb all that information.

[trh01uk]

David and Jeremy, thanks for the explanations of the circuit. I will add a few further points to ponder on.

The first is that actual modulation spec of the generator. The internal modulation is at 400 and 1,000c/s. The external modulation extends to 20kc/s and down to 20c/s, but the modulation depth and the max mod frequency are dependent on the carrier frequency selected - see the specification excerpt I have included. I think this is a mess - the spec for ext mod is so complicated as to be unusable.....(though it probably didn't matter too much for the original users)

Jeremy, you wrote:

If you wanted to AM modulate via the ALC you would surely inject the AM at the reference pin of the error amplifier and this is the 'SET ALC' pin3 of the V102 valve.

I don't agree that the "reference pin" of the error amplifier is pin 3 (screen grid) of V102. I think the reference is in fact the cathode, since it is primarily the cathode voltage that will be compared with the control grid input voltage.

I further suspect that the "SET ALC" control is really a "SET ALC GAIN" control. Varying the screen grid voltage as RV101 does should change the gain of V102. RV102 will change the set point by varying the cathode voltage - and indeed is labelled "SET CARRIER".

There seem to be quite a few attempts in this loop at controlling the frequency response. The principal one that I see is the series combination of R107 and R171 (1.1M total) followed by C160 (200pF) to ground plus at least the input capacitors of filters A and B. If we take the 0.01uF of filter A this gives a rolloff at about 15Hz, thus suggesting that modulation at practical frequencies is stripped out of the loop.


David, you wrote:

The anode potential of V1032 isn't reduced enough by the cathode follower, so the neon bulb V104 is used as an additional fixed voltage dropper to put the output DC in the right vicinity for the 5763's g2.

I think you are missing my point here. We have a series combination of R110 (10k), V104 (the neon), R108 (22k) and R185 (100R). According to the schematic the voltage at the R110 end is 44 volts. At the other end we have 25 volts. So we are dropping only 19V across this string - but every neon I have ever seen has dropped at least 90V before it passes any current. So either I am missing something here - or we have some typo on the schematic. If its a typo, I can't spot it. 44V at V102 (triode) cathode looks about right if its control grid has 36V on it. And 36V from the anode of the pentode section looks about right with an anode load of 1M.


Richard

[G0HZU_JMR]

Jeremy, you wrote:

If you wanted to AM modulate via the ALC you would surely inject the AM at the reference pin of the error amplifier and this is the 'SET ALC' pin3 of the V102 valve.

I don't agree that the "reference pin" of the error amplifier is pin 3 (screen grid) of V102. I think the reference is in fact the cathode, since it is primarily the cathode voltage that will be compared with the control grid input voltage.

I further suspect that the "SET ALC" control is really a "SET ALC GAIN" control. Varying the screen grid voltage as RV101 does should change the gain of V102. RV102 will change the set point by varying the cathode voltage - and indeed is labelled "SET CARRIER".

I have to hold my hand up and admit again that I really don't understand the way complicated valves work. When I suggested modulating pin 3 of the valve I was just guessing really I was trying to draw a parallel with how an opamp with + and - inputs might be configured here. Thanks for clarifying things a bit better, Richard.

I do think that the ALC detector works the way I suggest but I analysed it in isolation. i.e. I'm suggesting what would happen at C160 if you force fed the ALC detector an AM modulated signal from something like a sig gen. I don't think the DC voltage here at C160 will change regardless of modulation depth. But I think you would need a sig gen that can output quite a large RF envelope to mimic the TF144.

I would go further and suggest that you could feed it a different modulation waveform and the result would be the same assuming that the waveform is symmetric (or mirrored) around 0V. eg with a simple square or ramp waveform replacing the sine wave. Even if I'm right I'm not sure how useful this is to the current fault condition though

[G0HZU_JMR]

There seem to be quite a few attempts in this loop at controlling the frequency response. The principal one that I see is the series combination of R107 and R171 (1.1M total) followed by C160 (200pF) to ground plus at least the input capacitors of filters A and B. If we take the 0.01uF of filter A this gives a rolloff at about 15Hz, thus suggesting that modulation at practical frequencies is stripped out of the loop.

Yes, that's exactly as I see it.

I suspect that the loop will try and keep the average voltage at C160 (measured in slow time) to be the same with or without modulation present. Typical modulation frequencies will change the voltage at C160 quicker than the loop can respond so all the system can hope to do is keep the average voltage at C160 in the same place if/when the modulation depth changes across 0 to 80%.

[Radio Wrangler]

Thanks for the spec, Richard. Yuk! it's slew rate limited in the amplifier hence the dropping max mod depth with increasing mod frequency.

Jeremy is definitely right that this is a slow loop, and that the forwards gain of the amp with its local feedback is used to give simple gain to the mod signal heading for g2. So the mod depth will change with frequency and especially band.

VR104 is the only DC path to-from the 5763 g2 on ranges A-H and 44v down to 25v doesn't fit with any of the normal neon bulbs or reference tubes. There were some low voltage ones that never caught on due to being a bit radioactive and they're rarer than those thermocouples. But there's no R to ground to give sustaining current to keep the thing ionised unless the 5763 is in significant g2 current....

The whole thing is rather bodgy and has fixes on top of fixes.

David

[G0HZU_JMR]

It's certainly got lots of fudges fitted I deliberately avoided mentioning the strange thing that looks like a neon. What does that thing actually do I wonder?

One thing that concerns me wrt the repair and the symptoms of low DC voltage at C160 is that there are lots of capacitors in the modulation LPFs that could leak and drag down this DC voltage. So they would all need to be tested for leakage.

I do think you will need at least a high impedance DMM to measure the voltage here and I'd worry if the act of touching a DMM here would introduce local noise pickup into the system. This might even upset the readings.

If someone tells me a typical Vpkpk at the input to the ALC detector I think I can calculate the expected dc voltage at C160 to test if all the shunt leaks (if any) are gone. If this proves useful then my detector analysis will have been worthwhile I think. However, I think it might be wise to build one and test it to prove I've analysed it correctly. I think the detector will ride on top of the 2.5V DC pedestal that comes from the div/100 circuit from the 250V HT supply.